Search Type Title Author/Investigator Search Advanced Search

Table of Contents | Previous | Next

VII. DIFFERENTIAL EFFECTS OF EARLY HEAD START PROGRAMS ON
CHILDREN AND FAMILIES WITH DIFFERENT CHARACTERISTICS

Beyond examining impacts overall and in key subgroups of programs, it is important to look at variations in impacts among key subgroups of families. For whom did Early Head Start make a significant difference in outcomes? And how did the impacts differ among families? Variations in impacts might provide insights into how the programs influenced children and families and could identify demographic groups that merit special attention in future training and technical assistance.

In this chapter, we present impacts for selected key subgroups. Key tables are at the end of the chapter (p. 358). Additional subgroup analyses are presented in Appendix E tables. The subgroups we focus on here include subgroups based on whether the family enrolled before the child was born, age of mother at child’s birth, whether the child was the firstborn child, race/ethnicity, number of maternal risk factors, and for a subset of sites, whether the mother was at risk of depression when the family enrolled. In Appendix E, we present additional tables showing impacts for subgroups defined by other family characteristics, including the child’s gender and the primary caregiver’s living arrangements/marital status, receipt of welfare cash assistance, primary occupation (employment and school status), and highest grade completed at the time of enrollment. The subgroups highlighted in this chapter were selected because the patterns of impacts in these subgroups have the greatest implications for program practices1.

Our analyses of variations in impacts among family subgroups show that the Early Head Start research programs had significant impacts on some outcomes in almost every subgroup of families we studied, although the extent and pattern of impacts varied:

• The Early Head Start programs reached all types of families with child development services. They had significant positive impacts on service receipt in all subgroups of families we examined.

• By age 3, most groups of children benefited in some way from participating in Early Head Start. The programs had significant favorable impacts on at least one child development outcome for African American and Hispanic children, children who were enrolled prenatally and those enrolled after birth, firstborn and later-born children, children whose mothers lived with an adult other than their spouse, children of teenage and older mothers, children in families that were receiving TANF cash assistance and children in families that were not, children in all groups of families by primary occupation and highest grade completed, and children in families with fewer risk factors. A few groups of children did not benefit significantly, including children in white non-Hispanic families, children who lived alone with their mothers, children living with two parents, and children in the highest-risk families who enrolled (for whom the programs had significant negative impacts on some outcomes).

• Most parents benefited from participating in Early Head Start in some way related to their role as parents. Primary caregivers in all subgroups that we examined except one (those who were not receiving welfare cash assistance when they enrolled) experienced significant impacts on at least one aspect of parenting and family functioning by the time their child was 3 years old. Most subgroups experienced significant impacts on more than one aspect of parenting.

• Early Head Start also helped parents in most subgroups work toward economic self-sufficiency. The programs had positive impacts on participation in education and job training activities in all of the subgroups except families that enrolled with laterborn children, two-parent families, and lower-risk families. The programs also had positive impacts on employment in some of the subgroups of parents, including those who were not teenagers when their child was born, parents of firstborn children, non- Hispanic African Americans, mothers who were not receiving welfare cash assistance when they enrolled, parents who were neither in school nor employed when they enrolled, and parents who had completed high school.

• The programs significantly delayed subsequent births in several subgroups. Although delaying subsequent births was not a goal of Early Head Start, programs worked with families toward their goals, which may have included delaying subsequent births, and made referrals to health care and family planning providers. Program participation led to significant delays in subsequent births among Hispanic and non-Hispanic, white families; families who enrolled with firstborn children; mothers who lived alone with their children; mothers who were receiving welfare cash assistance when they enrolled; mothers who were in school or neither employed nor in school; mothers who had not yet completed high school; and the highest-risk families.

Below, we highlight important variations in program impacts among key family subgroups. Because of the large number of subgroups and outcomes, we focus primarily on patterns of impacts. In the next section we present the hypotheses that guided our choice of subgroups and expected differences in impacts, describe our approach to estimating and interpreting subgroup impacts, and highlight variations in impacts across key subgroups. In the following section we highlight the estimated program impacts for several key policy-relevant subgroups and discuss their importance. The chapter ends with a discussion of the implications of these findings.

A. IMPORTANT VARIATIONS IN PROGRAM IMPACTS FOR CHILDREN AND FAMILIES WITH DIFFERENT CHARACTERISTICS

Our investigation and interpretation of differences in impacts among subgroups of families was guided by the hypotheses that are discussed in the first subsection below. The next subsection provides a brief overview of our approach to estimating subgroup impacts and conducting analyses to help interpret them. The following subsections present the analysis findings for key subgroups.

1. Guiding Hypotheses

Child’s Age at Enrollment. Impacts may differ among families in which the mother enrolled while pregnant and families in which the mother enrolled during the child’s first year of life because the duration of program participation is potentially longer (by as much as 15 months) among those who enrolled before the child was born. Among program group families, those who enrolled while pregnant remained enrolled for an average of 25 months, while those who enrolled after their child was born remained enrolled for an average of 22 months. At each assessment point (2 and 3 years of age), the children who were enrolled prenatally had greater exposure, on average, to program services than children who were enrolled after birth.

Opportunities for improving child outcomes may also be maximized when program staff begin working with families prenatally and ensure that pregnant women receive prenatal care and education (Olds, Henderson, Kitzman, Eckenrode, Cole, and Tatelbaum 1999). Moreover, pregnancy may be a time when parents are more open to intervention services as they work through changes in their lives (Duncan and Markman 1988; Brazelton and Cramer 1991; Osofsky and Culp 1993).

Birth Order. Opportunities for changing parenting behavior and improving child outcomes may be maximized when program staff begin working with first-time parents who may be feeling uncertain about their new roles as parents and most receptive to program guidance related to parenting (Olds, Henderson, Kitzman, Eckenrode, Cole, and Tatelbaum 1999).

Impacts may be smaller among families of later-born children if they have established patterns of parenting behaviors with earlier children that are difficult to change. On the other hand, impacts may be larger if parents enrolling with later-born children have faced challenging parenting experiences in the past and therefore value help with parenting more than first-time parents, if the program helps parents with several children to pay special attention to their infant, or if direct services to children compensate for limited attention from parents with several children.

Age of Mother When Child Was Born. Teenage mothers are likely to be less emotionally mature than older mothers, and they may be struggling with their own developmental needs and less receptive to some services directed toward their children’s development (Wakschlag et al. 1996; Moore, Brooks-Gunn, and Chase Lansdale in press; Chase-Lansdale and Brooks-Gunn 1994). Perhaps because they are often less emotionally mature, program staff regarded teenage mothers as harder to serve. Staff rated fewer teenage parents as consistently highly involved in the program (30 percent compared with 40 percent of older mothers). Thus, impacts may be smaller among teenage parents.

On the other hand, because teenage parents and their children face higher risks for poor outcomes than older mothers (see for example, Maynard 1996), those whom the programs are able to engage in services may benefit more. Also, center-based child development services might help teenage mothers stay in school and enhance children’s cognitive development (Brooks-Gunn, Fuligni, and Berlin 2000).

Because teenage parents and their children face higher risks of poor outcomes, they are often the targets of intervention programs. If teenage parents in the control group were more likely than older mothers in the control group to obtain similar services, Early Head Start impacts on teenage parents and their children might be smaller than those for older parents.

Race/Ethnicity. Impacts may differ among racial/ethnic groups because of cultural differences affecting families’ receptiveness to formal support services, and in the case of Hispanic families, language barriers that may interfere with services, especially services and resources to which Early Head Start refers them in the community. The impacts may also differ because pre-existing cultural practices or attitudes related to parenting or child development may interact in unique ways with program services. Early Head Start programs are expected to provide services that meet families’ needs and are given wide latitude for designing services that are culturally appropriate.

Nevertheless, families from different cultural backgrounds may experience and respond to various Early Head Start services differently. The average duration of Early Head Start enrollment was slightly longer among African American families (23.3 compared with 21.9 and 22.9 months in white and Hispanic families), and African American and Hispanic families were more likely than white families to remain enrolled for two years or longer (55 and 58 percent compared with 48 percent). On the other hand, program staff were more likely to rate Hispanic and white families as consistently highly involved in the program (41 and 38 percent compared with 32 percent). These variations in the duration and level of program involvement may contribute to differences in program impacts.

Cultural biases in child and parenting outcome measures could also contribute to variations in impacts by race/ethnicity. We attempted to minimize these biases by choosing measures that had previously been shown to work well in varied racial and ethnic groups. In addition, as we examined the psychometric properties of the child and family measures, we calculated internal consistency alphas for each of the three major racial/ethnic subgroups. For the most part, the measures appeared to be appropriate for all groups of children and families. Nevertheless, it is possible that cultural biases could affect the measures in other ways.

Number of Risk Factors. All Early Head Start families are at risk of poor outcomes due to poverty. Some are at greater risk than others, however. In order to distinguish families with different levels of risk, we counted the number of demographic risk factors that families had when they enrolled (in addition to being low income, a characteristic that most Early Head Start families shared). Some of the risk factors tended to occur together, and when they did, families were considered higher-risk families. We counted up to five risk factors: (1) being a single parent; (2) receiving public assistance; (3) being neither employed nor in school or job training; (4) being a teenage parent; and (5) lacking a high school diploma or GED. To form subgroups of reasonable size, we divided families into three groups based on the number of risk factors they had when they enrolled: (1) families who had zero, one, or two risk factors; (2) families who had three risk factors; and (3) families who had four or five risk factors.

Impacts among families with varying numbers of risk factors may differ for two possible reasons. First, program staff reported that it was harder to engage and serve higher-risk families, and they often found it necessary to address critical economic and social support needs before parents in this group were able to focus on child development services. The challenges of serving families with more risk factors are reflected in lower average durations of program enrollment, a lower likelihood that they remained enrolled for at least two years, and smaller percentages rated by staff as consistently highly engaged in the program. For that reason, program impacts on service use, especially intensive service use, may be smaller among families with more risks, and as a result, child and parenting outcomes might also be smaller among these families. Second, in the control group, families with more risks may have had more difficulty than families with fewer risks with obtaining similar services in the community. For that reason, impacts might be larger among families with more risks.

On balance, impacts on families with more risks may be smaller or larger than those on families with fewer risks. The evaluation of the Infant Health and Development Program found that among children in poor families, the effects of the intervention were largest for those with low or moderate risks, and there was no impact on cognitive development when risks were high (Liaw and Brooks-Gunn 1994).

Maternal Risk of Depression. For 8 of the 17 research programs, data on depressive symptoms were collected at the time of enrollment. Mothers who reported depressive symptoms and were at risk of depression when they enrolled may have been struggling with their own mental health needs and less receptive to some services directed toward their child’s development. Program staff also regarded mothers with mental health needs as harder to serve. Thus, we might expect smaller impacts on the parenting and child development outcomes among families of depressed mothers. On the other hand, mothers in the control group who were at risk of depression may have been less likely than control-group mothers who were not at risk of depression to seek other services, and the Early Head Start programs may have had a greater opportunity to have an impact on parenting and child outcomes among families of mothers at risk of depression.

2. Approach to Estimating and Interpreting Subgroup Impacts

Our basic approach to estimating subgroup impacts was to average site impacts across sites where there were at least 10 program and 10 control group families in the subgroup. When this strategy resulted in several sites being omitted from some subgroups, we tested the sensitivity of the findings to this assumption by pooling data across sites and using all available observations from all sites to estimate impacts2.

Caution must be used in interpreting the variations in impacts among subgroups of families. The subgroups are defined on the basis of a single family characteristic, yet they may also differ in other characteristics. These other unaccounted-for variations in family characteristics may also influence the variations in impacts. Thus, in our analyses we focus on patterns of impacts across outcomes and consider the potential role of other differences in characteristics that may have influenced the outcomes examined. We also conducted analyses in which we controlled for multiple characteristics simultaneously to help assess the extent to which confounding of characteristics may account for the results from the basic approach.3 However, these analyses cannot control for differences in unmeasured characteristics and it is not possible to rule out all potential sources of confounding.

In discussing the subgroup findings below, we focus on several different aspects of the findings. We compare impacts across family subgroups and focus primarily on those differences in impacts that are statistically significant. We also discuss impacts within particular subgroups that are statistically significant or relatively large (in terms of effect sizes). Some of the family subgroups are small and power to detect significant differences is low. In these subgroups, especially, we note relatively larger impacts even when they are not statistically significant in order to identify patterns of findings. In drawing conclusions from the impact estimates, we focus on patterns of impacts across outcomes.

3. Variations in Impacts By Mother’s Pregnancy Status at Enrollment

Impacts on Service Use. Impacts on service use among families in which the mother enrolled while pregnant with the focus child tended to be larger than those among families in which the mother enrolled after the focus child was born (see Table VII.1 at the end of the chapter). This generally reflects higher rates of service receipt by families in the program group who were pregnant when they enrolled.

The impacts on receipt of intensive services also tended to be larger among families who enrolled while pregnant. One exception to this pattern is in the area of child care services, where the impacts on average hours per week in center-based child care and average weekly out-ofpocket child care costs were larger among families who enrolled after the focus child was born. This likely reflects the fact that pregnant women did not need child care services during the early portion of the follow-up period and were more likely to be receiving home-based services initially.

Impacts on Child and Family Outcomes. Early Head Start had a favorable impact on the cognitive and language development and social-emotional behavior of 3-year-old children whose mothers entered the program while pregnant and those who entered during their first year of life, but the impacts tended to be greater for children whose mothers entered during pregnancy (Table VII.2). Impacts on average Bayley MDI scores were positive and statistically significant among children whose mothers entered during pregnancy. Some impacts on positive social-emotional behavior were favorable and statistically significant for both subgroups, but they were often larger for children whose mothers entered Early Head Start during pregnancy. Early Head Start participation led to a significant reduction in the children’s sustained attention with objects and engagement of their parents during semi-structured play for both subgroups, but the impacts were larger for children whose mothers entered the program during pregnancy. In addition, the programs had significant favorable impacts on children’s negativity toward their parents, children’s engagement of their parents in the puzzle challenge task, and persistence in the puzzle challenge task among children whose mothers enrolled during pregnancy.

For some aspects of parenting behavior, the impacts of Early Head Start were larger among mothers who entered during pregnancy, while for other aspects of parenting behavior, the impacts were larger among mothers who entered during their child’s first year of life. Impacts on the overall organization, emotional support, and support for cognitive development of the home were favorable for both groups of parents, but were statistically significant only for families entering during the child’s first year of life. Impacts on the parent’s stimulation of language and learning were generally favorable and sometimes statistically significant for parents entering the program in the child’s first year of life, but were not statistically significant (and not always favorable) for parents entering during pregnancy. Impacts on emotionallysupportive parenting behavior, while positive and statistically significant for both groups, were often larger for parents entering Early Head Start during pregnancy. Early Head Start tended to reduce negative parenting behavior among both groups of parents, but the subgroup impacts in most cases were not statistically significant. Early Head Start reduced spanking more among parents who enrolled during pregnancy than those who enrolled after their child was born.

When the children were 3 years old, Early Head Start participation led to higher rates of self-reported symptoms of depression among mothers who entered the program during pregnancy. A similar impact on depression was not found when children were 2 years old, however, suggesting that families who enrolled during pregnancy and participated in Early Head Start until their children were 3 years old may have been experiencing some distress associated with transitioning out of Early Head Start.4 Impacts on symptoms of depression were favorable for parents entering Early Head Start in the child’s first year of life, but not statistically significant.

Early Head Start led to greater participation in self-sufficiency activities among parents in both groups (Table VII.3). The favorable impacts on overall participation in education and training programs were statistically significant for both groups of parents. The impacts over time were more consistent among parents who enrolled during their child’s first year of life. The impacts on quarterly participation in education programs among these parents were consistently positive and statistically significant beginning in the third quarter after enrollment and extending throughout the remaining follow-up period.

The somewhat stronger pattern of impacts in most areas among families that enrolled while pregnant is consistent with the longer duration of services they received and their potentially greater receptiveness to services. This pattern of impacts suggests that it may be advantageous to enroll families prenatally when possible.

It is important to note, however, that the Early Head Start programs also had significant favorable impacts on children and parents who enrolled after their child was born. The results suggest that it is not too late to make a difference after the child is born.

The differences in impacts when children were 3 years of age between families who enrolled during pregnancy and families who enrolled after the child was born tended to be less consistent across outcomes than they were when children were 2 years of age. Over time, the difference in potential exposure to program services appears to have made less of a difference in program impacts.

4. Variations by Child’s Birth Order

Impacts on Service Use. Impacts on service use and receipt of intensive services tended to be larger among families in which the focus child was not the firstborn child (Table VII.4). One exception is that the impacts on use of any child care and use of center-based child care were larger among families who enrolled with a firstborn child (although the impact on average hours per week of center-based care was virtually the same in the two groups).

Impacts on Child and Family Outcomes. The favorable Early Head Start impacts on children’s cognitive and language development did not differ significantly among firstborn and later-born children (Table VII.5). Most impacts on children’s social-emotional behavior also did not differ significantly, but the favorable impact on children’s engagement of their parents during play was significantly larger among firstborn children. The patterns of impacts on child outcomes are similar to those observed when children were 2 years old.

Similarly, Early Head Start tended to have favorable impacts on the parenting behavior of parents who entered the program with firstborn and parents who enrolled with later-born children. Impacts were more often statistically significant for parents of firstborn children, but this subgroup was somewhat larger than the subgroup of parents with later-born children. Early Head Start impacts on discipline were significant and much larger among parents who enrolled with later-born children. Early Head Start had no significant impacts on the self-reported mental health of parents who entered the program with either firstborn children or those who enrolled with later-born children.

Early Head Start boosted participation of parents in self-sufficiency activities, but the pattern of activities affected varied across the groups (Table VII.6). The Early Head Start programs increased participation by parents of firstborn children in education activities overall and consistently increased the participation of parents of firstborn children in educational activities significantly in the third through eighth quarters after enrollment. Early Head Start more consistently increased employment rates among parents of later-born children. Parents of laterborn children participating in Early Head Start were more likely than similar control group parents to be employed, especially in the earlier quarters of the follow-up period. The programs also significantly reduced the proportion of parents of firstborn children who had another birth during the first two years after enrollment.

Confounding with other factors does not appear to account for the patterns of findings described above. The patterns of impacts among families who enrolled with firstborn and laterborn children are similar when other factors are controlled simultaneously in multivariate models. These models continue to show that the programs had favorable impacts on both groups of families. Although we expected to find larger impacts among firstborn children and their 332 parents, the evaluation findings support the value of intervention for both firstborn and later-born children.

5. Variations in Impacts Among Teenage and Older Mothers

Impacts on Service Use. Program impacts on service use and on intensity of services received were consistently larger among older mothers than teenage mothers (Table VII.7). For many types of services, teenage mothers in the control group were more likely than older mothers in the control group to receive services and to receive intensive services, reflecting the availability of supportive services for teenage parents in many communities. At the same time, service receipt, particularly receipt of intensive services, by teenage mothers in the program group tended to be lower than service receipt by older mothers in the program, consistent with staff perceptions that it was harder to serve teenage mothers. The only exception was child care use by teenage mothers in the program group, which was generally higher than child care use by older mothers in the program group.

Impacts on Child and Family Outcomes. The Early Head Start impacts on the average levels of cognitive development of 3-year-old children did not differ significantly between children of teenage and older mothers. Early Head Start participation, however, raised the proportion of children of teenage parents who received Bayley MDI scores above the threshold score of 85 by a significantly greater amount (Table VII.8). In the control group, teenage mothers were much more likely than older mothers to have children who received Bayley MDI scores below 85; Early Head Start participation led to reductions in the proportion of children of teenage mothers who received low scores to the level found among older mothers. The program significantly improved the language development of children of older mothers, but had no statistically significant impacts on the language development of children of teenage mothers.

Early Head Start had favorable impacts on the social-emotional behavior of children of both teenage and older mothers. Impacts on engagement of the parent in play were positive and significant for both groups of children. The impact of Early Head Start on sustained attention to objects during play was significantly greater for children of teenage parents than for children of older parents. Early Head Start reduced negativity toward the parent in play and aggressive behavior problems among children of older mothers. The impacts on negativity and aggression among children of teenage mothers were favorable and not statistically different from the impacts for older mothers, but they were not large enough to be statistically significant.

Early Head Start had favorable impacts on a broad set of measures of parenting behavior for older mothers, but also had significant impacts on the parenting behavior of teenage mothers in a few areas (supportiveness and discipline). Scores on the HOME were significantly increased among older mothers participating in Early Head Start. Supportiveness during parent-child play was enhanced significantly for both teenage and older mothers. Parent stimulation of the child’s language development and learning, including daily reading, was generally enhanced for older mothers, but no impacts were detected for teenage mothers. Early Head Start generally had no significant impacts on negative parenting behavior for either teenage or older mothers, with one exception. The proportion of parents who reported using physical punishment in the past week was significantly lower for both teenage and older parents, and the use of physical punishment as a discipline strategy tended to be lower for both groups. The pattern of impacts on parenting outcomes among older mothers was stronger when children were 3 years old than when they were 2 years old.

Early Head Start had no impacts on the mental health of either teenage or older parents when children were 3 years old. The significant reductions in parental distress and dysfunctional 334 parent-child interactions found among teenage parents when children were 2 years old did not persist.

Early Head Start led to greater participation in self-sufficiency activities by both teenage and older parents (Table VII.9). Early Head Start increased the likelihood that parents participated in education programs, increasing the enrollment of teenage mothers in high school programs and increasing the enrollment of older mothers vocational education programs. Early Head Start also increased employment rates among older mothers but had no significant impact on the employment of teenage mothers.

These findings reflect the emphasis Early Head Start programs tended to place on pursuing education so that parents might qualify for higher-wage jobs with fringe benefits. Education was a goal particularly for parents who had not finished high school, many of whom were teenage parents. It is notable that Early Head Start increased participation in education programs among teenage parents, even when control-group participation was high, probably because organizations in many communities also support education for teenage parents, and new requirements for welfare recipients mandate school attendance for unmarried parents under 18 years old. Although the Early Head Start programs increased participation rates in education programs among teenage parents, they did not significantly increased the proportion of teenage parents who had completed a high school degree or GED by two years after enrollment.

Initially, the Early Head Start programs increased welfare receipt among teenage mothers, but by the last two quarters of the follow-up period, the programs had begun to reduce welfare receipt among teenage parents significantly. The programs did not have a significant impact on welfare receipt among older mothers.

Confounding with other factors does not appear to account for these patterns of impacts. The estimated impacts are similar when other factors are controlled. The weaker pattern of impacts on child development and parenting among teenage parents and their children supports the hypothesis that teenage mothers were less mature and less receptive than older mothers to services directed toward their children’s development.

6. Variations in Impacts by Race/Ethnicity and Language

We examined impacts for three racial/ethnic groups: non-Hispanic, African American families; Hispanic families; and white, non-Hispanic families. The numbers of families in other racial/ethnic groups were too small to examine impacts for them separately. Because language differences may be related to cultural differences and help us understand the differences in impacts among racial and ethnic groups, we also examined impacts for families whose primary language was English and families whose primary language was not English (usually Spanish).

Impacts on Service Use. Impacts on service use were large and significant in all racial/ethnic groups (Table VII.10). Impacts on use of any services by Hispanic families by 28 months after enrollment tended to be much larger than for other families, primarily because Hispanic control group families were much less likely than other control group families to receive services.

Impacts on receipt of intensive services—core child development services at the required intensity, weekly home visits, and weekly case management—during the 28-month follow-up period were largest among white, non-Hispanic families, primarily because service receipt by program group members was highest among white families. However, impacts on average hours of center care per week were largest for Hispanic families and families whose primary language was English (Table VII.13).

Impacts on receipt of weekly home visits were larger among English-speaking families in the first follow-up period, but larger among non-English-speaking families in the second and 336 third follow-up periods. The impacts on receipt of weekly home visits in at least one follow-up period and in all three follow-up periods were similar in the two groups.

Impacts on use of services and receipt of intensive services by African American families by 28 months after enrollment tended to be smaller than the impacts for other families. This pattern often reflects relatively higher levels of service use among African American control group members as well as relatively lower levels of service use among African American program families. However, the impacts on child care use by African American families, while smaller than those for Hispanic families, were larger and more often significant than those for white, non-Hispanic families. Levels of child care use tended to be highest among African American families in both groups relative to their counterparts among Hispanic and white families.

Impacts on Child and Family Outcomes. The Early Head Start impacts on average levels of cognitive and language development did not differ significantly among families of different racial and ethnic backgrounds. Although the impacts in individual racial/ethnic groups were not statistically significant, Early Head Start had a significantly more favorable impact on the proportion of children scoring below 85 on the Bayley MDI among Hispanic households and households in which the primary language was not English (Table VII.14). The impact on the average PPVT score was positive and significant for African American children. Although it was not statistically significant, the reduction in the proportion of children who scored below 85 on the PPVT-III was significantly greater among African American children. Similar impacts on language outcomes were found when children were 2 years old.

Early Head Start appears to have improved language development among Hispanic children as well. The impact on the average PPVT score was positive but not statistically significant for Hispanics because some children in this group completed the PPVT and some completed the TVIP.5 The impact of Early Head Start on TVIP scores was also positive but not statistically significant for Hispanic children. The positive trend in the scores on both assessments suggests that overall, it is likely that Early Head Start improved language development for Hispanic children. These potential positive impacts on language outcomes among Hispanic children were not apparent at the earlier assessment.

Early Head Start had statistically significant, favorable impacts on the social-emotional behavior of 36-month-old African American children, while the impacts on the behavior of white or Hispanic children were not significant. Among African American children, Early Head Start participation led to reduced aggressive behavior and child negativity toward the parent in a semistructured play task, enhanced children’s sustained attention with objects and engagement of the parent in the play task, and increased children’s engagement of their parents and persistence in a puzzle challenge task. The impacts on African American children were more consistent and larger than those seen when the children were 2 years old. The few significant impacts on white children’s social-emotional behavior observed at 2 years of age did not persist when the children were 3 years old.

The impacts of Early Head Start on parenting when children were 3 years old are generally consistent with the impacts on children’s development and behavior. Early Head Start enhanced emotionally-supportive parenting among African American parents and reduced intrusiveness during semi-structured play and during a puzzle challenge task among African American parents. Impacts were negligible for white and Hispanic parents. The favorable impacts on emotionally supportive parenting and the reduction of negative parenting behavior among African American parents may partly explain the favorable impacts on African American children’s behavioral outcomes.

Program impacts on parents’ stimulation of language and learning were significantly greater among both African American and Hispanic parents, and the programs increased the percentage of Hispanic parents who reported reading to their children daily as well. These impacts may partly explain the favorable impacts on cognitive and language development for African American and Hispanic children.

The range and size of Early Head Start impacts on parenting among African American families increased over time. More impacts on parenting were significant, and impacts tended to be larger when children were 3 years old. The impacts on parenting observed among white families when children were 2 years old did not persist when children were 3 years old.

Early Head Start improved aspects of mental health among African American parents, but appears to have had unfavorable impacts on the mental health of white parents. Parental distress and parent-child dysfunctional interaction were significantly reduced among African American parents participating in Early Head Start, while Early Head Start appears to have increased parent-child dysfunctional interaction among participating white parents.

Patterns of program impacts on self-sufficiency activities varied among the racial/ethnic groups (Table VII.12). The Early Head Start programs increased the proportion of African American parents who were employed at some time during the two-year follow-up period, but in the final two quarters of the follow-up period, positive impacts on participation in education activities also emerged. Early Head Start increased participation in education activities and reduced employment among Hispanic parents early in the follow-up period, but later in the follow-up period, the impacts on participation in education activities faded and positive impacts on employment emerged. Among white families, Early Head Start led to an increase in participation in education programs, particularly in the second year of follow-up, but had no significant impacts on employment. Program participation led to a significant reduction in subsequent births during the two years after enrollment among white and Hispanic families.

Early Head Start increased the receipt of TANF cash assistance significantly among Hispanic families but not among the other groups of families. Among control families, levels of TANF receipt were much lower among Hispanic than other groups of families. The programs brought the levels of TANF receipt among Hispanic families closer to the levels for program families in other racial/ethnic groups, but they remained lower. It appears that the Early Head Start programs helped some eligible Hispanic families who may have had reservations about seeking cash assistance or had language barriers to obtain the assistance they needed.

The notably strong favorable pattern of impacts for African American families, the pattern of favorable impacts for Hispanic families, and the lack of significant impacts among white non- Hispanic families persist when impacts are estimated by pooling data across sites and eliminating the requirement that there be ten program and ten control families in the subgroup for a site to be included in the analysis.6

To the extent that it is possible to investigate, confounding with other characteristics does not appear to explain the pattern of impacts by race/ethnicity. African American families were more likely to be served in mixed-approach programs, and the parents were more likely to be teenage mothers in school or training who entered the programs with firstborn children. Hispanic families were less likely to speak English as their primary language, less likely to have completed high school or a GED, older, more likely to enroll with later-born children, and more likely to be served in programs that were fully implemented later. White families were more likely to be lower-risk families served in home-based programs and programs that were early implementers. However, when we estimated multivariate models controlling simultaneously for multiple site and family characteristics, the pattern of impacts by race/ethnicity persisted. Nevertheless, it is possible that confounding with other unmeasured characteristics may explain the differences in impacts by race/ethnicity.

The status of African American control group children and families relative to the control families for other racial/ethnic groups may have set the stage for the Early Head Start programs to make a larger difference in the lives of the African American children and parents they served. For example, African American control group children had lower Bayley MDI scores than either Hispanic or white children, and lower PPVT-III scores than white children. Non-Hispanic white children in the control group tended to be in a more favorable position than African American and Hispanic children in the control group (Table VII.11). Similarly, non-Hispanic white parents in the control group tended to demonstrate the most favorable parenting behaviors and African American parents in the control group tended to demonstrate the least favorable parenting behaviors, and in some cases the differences among the racial/ethnic groups were large.

Although the impacts on service use tended to be smaller among African American families, because control group families were more likely to receive services, the services received by the control group families were less likely to be intensive and may not have been as effective as those provided by Early Head Start. In particular, levels of child care use, including use of center-based care, were relatively high among African American families in both the program and control groups. Differences in the quality of child care used by the two groups may have contributed to the larger impacts on child development outcomes in this group.

It is notable that white, non-Hispanic families in the control group were more likely than other control group families to report that their child was eligible for early intervention services and more likely to receive such services, suggesting that white children in this sample may have been more likely to have a disability. It is likely that the early intervention services received by some white control group families and children were comprehensive and in many ways similar to the Early Head Start services received by program families.

In both the program and control groups, the parents in white families received higher scores on the CES-D and were more likely to be experiencing moderate or severe depression when their children were 3 years old than African American and Hispanic parents. These differences were apparent when children were 2 years old as well. The higher incidence of depression among white parents may have contributed to greater challenges for programs in serving white families and less success in achieving impacts with them.

The evaluation of the Infant Health and Development Program (IHDP) also found stronger effects for African American families (Brooks-Gunn et al. 1993). However, these stronger effects were due largely to differences in education. IHDP’s effects on cognitive and language development when children were 3 years old were found for children of African American and white mothers with less than a high school education and for those with a high school diploma but no more, but not for those with more than a high school education (Brooks-Gunn et al. 1992). Almost no early studies of similar programs included a sufficient number of white families to allow comparisons of impacts by race-ethnicity. First and second generation evaluation studies included mostly African American families, with a few Hispanic families also included.

7. Variations in Impacts By Number of Demographic Risk Factors

As noted earlier, we examined variations in impacts by the number of demographic risk factors by dividing the sample into three subgroups: (1) families with zero to two risk factors; (2) families with three risk factors; and (3) families with four or five risk factors).

Impacts on Service Use. Impacts on service use and receipt of intensive services tended to be larger among families with fewer than three demographic risk factors (Table VII.16). This often reflects higher levels of service use by program families with fewer risk factors, compared with program families with more risk factors, consistent with program staff perceptions that higher-risk families were harder to serve. Impacts on child care use were similar among lower- and higher-risk families. The estimated impacts on receipt of core child development services at the required intensity throughout the follow-up period and on the use and intensity of centerbased child care were notably smaller among the small group of families with four or five risk factors.

Impacts on Child and Family Outcomes. Early Head Start impacts on the cognitive and language development and behavior of 3-year-old children differed significantly among families with different numbers of risks (Table VII.17). Children in families with two or three risk factors experienced a significant favorable impact on their Bayley MDI scores. Children in families with fewer risk factors experienced the greatest reduction in the proportion with PPVTIII scores below 85. The impacts of Early Head Start on the cognitive and language development of children in the families with more than three risk factors, however, were unfavorable. The impact on average PPVT-III scores was negative and statistically significant.

The estimated program impacts on children’s social-emotional behavior often did not differ significantly among the groups of families with different numbers of risk factors. Children in the highest-risk families, however, appeared to be unfavorably affected by Early Head Start participation. The impacts on orientation and engagement during the Bayley assessment and persistence and frustration in the puzzle challenge task were unfavorable among families with four or five risk factors. The unfavorable pattern of impacts that was found among this group of families when children were 2 years old persisted when they were 3 years old.

The favorable impacts of the Early Head Start programs on parenting were concentrated among families with three risk factors. Early Head Start had no statistically significant impacts on most parenting outcomes for the families with zero to two risk factors, except for a reduction in the use of physical punishment. Early Head Start had favorable pattern of impacts on parenting outcomes among families with three risk factors, including favorable impacts on outcomes in the areas of parents’ emotional support, stimulation of language and learning, negative parenting behaviors, and parents’ mental health. The Early Head Start programs had almost no statistically significant impacts on parenting among the parents in families with more than three risks, however, and the impact that was significant was an unfavorable impact on harshness toward the child during the parent interview. A few other impacts on parenting among these parents were relatively large and unfavorable. Again, this pattern is similar to that found when children were 2 years old.

Early Head Start led to a trend toward higher levels of parent-child dysfunctional interaction among parents with less than three risk factors. However, Early Head Start significantly reduced levels of parental distress among families with three risk factors.

Early Head Start had no consistent impacts on participation in self-sufficiency activities by parents with zero to two risk factors (Table VII.18). Early Head Start led to greater participation in education programs by parents with three risk factors. Among the families with four or five risk factors, the Early Head Start programs significantly increased welfare receipt, especially early in the follow-up period. It appears that the programs helped families who needed cash assistance obtain it.7 Participation in Early Head Start led to significant reductions in subsequent births during the two years after enrollment among the families with four or five risk factors.

The findings suggest that the program was most successful in improving outcomes among families who were in the middle of the range of number demographic risk factors. The unfavorable impacts among the small group of families with four or five risk factors suggests that the services provided by Early Head Start programs may not be sufficient to meet the needs of the families at greatest risk and may not be as effective as other community programs that target these families. The difficulties program staff reported in working with these families may be reflected in the less-favorable outcomes. In addition, the families with the most risks were more likely to be in home-based or mixed-approach programs that were not fully implemented early, and it is possible that the staff turnover and disruptions in staff-family relationships experienced in some of these programs had an adverse effect on the most vulnerable families.

8. Variations in Impacts By Mothers’ Mental Health Status

For these analyses, we focused on a subsample of eight programs for which data on parents’ feelings of depression were collected at enrollment. Parents were classified as at risk for depression at enrollment if they scored 16 or greater on the CES-Depression scale.

The eight programs for which data were collected on depressive symptoms at baseline included proportionately more mixed-approach programs and proportionately fewer center-based programs than the full sample. The eight programs also included proportionately more early implementers and proportionately fewer later and incomplete implementers. The families served by the eight programs with baseline data on depressive symptoms were similar to the full sample of families on some dimensions, but they were more likely to be white and less likely to be African American or Hispanic; more likely to enroll prenatally; less likely to be teenage mothers; and more likely to have completed high school or a GED. In these sites, approximately half of mothers were at risk of depression when they enrolled.

Impacts on Service Use. Impacts on overall service use were similar among those at risk and not at risk of depression when they enrolled (Table VII.20). However, impacts on intensive service use tended to be larger among families in which the mother was not at risk of depression. These larger impacts among families not at risk of depression reflect both less receipt of intensive services among control group families and greater receipt of intensive services among program families in this subgroup.

The programs increased the use of any child care significantly only among families in which the mother was not at risk of depression, but they increased use of center-based care in both groups, and increased use of any center-based care more among families in which the mother was at risk of depression (although the impact on average hours per week of center care was higher among families with mothers who were not at risk of depression).

Impacts on Child and Family Outcomes. The impacts of Early Head Start on cognitive development were not significant in either group, and they did not differ significantly between children with mothers at risk for depression at baseline and children with mothers not at risk (Table VII.20). The impacts on average language scores also were not significant in either group, but there was a trend toward a larger program-control difference for mothers not at risk compared with those who were at risk for depression. However, the Early Head Start programs significantly reduced the proportion of children scoring below the threshold of 85 on the PPVT III among children of mothers who were not at risk of depression at enrollment but not among children of mothers at risk.

Early Head Start had a consistent pattern of favorable, statistically significant impacts on the social-emotional behavior of children whose mothers were at risk for depression at enrollment but not among children whose mothers were not at risk. Program impacts on children’s engagement of their parents in both play and the puzzle challenge task, persistence in the puzzle challenge task, sustained attention with objects in play, and negativity toward their parents in play were all significant for children of mothers at risk for depression. The impact on child engagement of the parent in the puzzle challenge task was significantly greater than that for children of mothers who were not at risk for depression. The poorer social-emotional behavior of children of control group parents at risk of depression compared with children of control group parents not at risk of depression may have provided a greater opportunity for the programs to have a larger impact on this group of children.

Among parents not at risk of depression at enrollment, the Early Head Start impacts on parenting behavior were mixed. The impacts on emotionally-supportive parenting and most measures of support for language and learning were not significant. However, Early Head Start increased the proportion of parents who reported reading daily to their child more among parents who were not at risk of depression. Early Head Start tended to increase negative parenting behavior during the semi-structured play and puzzle challenge tasks among parents who were not at risk of depression, and the increase in negative regard during play was significant. However, Early Head Start tended to reduce the use of physical punishment among this group of parents.

Early Head Start had some notable statistically significant impacts on parenting behavior of parents at risk for depression at baseline, including significant increases in supportiveness in play and significant reductions in detachment and negative regard during play. Early Head Start also reduced spanking and reduced the severity of discipline that mothers who were at risk of depression reported they would use. The programs also increased the extent to which mothers who were at risk of depression reported following a bedtime routine with their child.

The estimated impacts on parent mental health were mixed among mothers who were at risk of depression at enrollment. Early Head Start significantly increased parent-child dysfunctional interaction among mothers at risk of depression, but also significantly reduced reported depressive symptoms among mothers in this group.

Early Head Start had no consistent impacts on self-sufficiency activities of parents at risk for depression at enrollment (Table VII.21). Among parents who were not at risk of depression at enrollment, Early Head Start increased participation in education and job training. The programs also increased employment in three out of the eight quarters following enrollment among these families.

Although the impacts on the receipt of intensive services were often smaller among families of mothers who were at risk of depression, the impacts on service receipt overall were similar among the two groups of families. The poorer outcomes among control group families in which the mother was at risk of depression at enrollment relative to other control group families in some areas, especially negative parenting behaviors, parent supportiveness, and children’s social-emotional development, may have set the stage for the Early Head Start programs to make a larger difference in these areas among families with mothers who were at risk of depression.

9. Other Subgroups Examined

We examined variations in impacts for several other types of subgroups, listed below, but do not discuss the findings here. Tables presenting the impacts for these subgroups are included in Appendix E.

• Subgroups based on receipt of welfare cash assistance at enrollment. We do not highlight these findings here because different rules for receiving cash assistance 348 were in effect for many of the families in the sample when they enrolled. Also, few differences in impacts on parenting and child development emerged in these subgroups (Appendix Tables E.VII.2 through E.VII.4).

• Subgroups based on child’s gender. We do not highlight these findings because the differences in child impacts that appeared when children were approximately 2 years old diminished or disappeared by the time they were 3 years old (Appendix Tables E.VII.5 through E.VII.7).

• Subgroups defined by parents’ primary occupation when they enrolled (employed, in school or job training, or neither). We do not highlight these findings here because they are generally similar to those for subgroups by number of maternal risk factors (being neither employed nor in school or training is one of the risk factors counted). They suggest that impacts were smaller for the families with the highest and lowest levels of education (Appendix Tables E.VII.8 through E.VII.10).

• Subgroups based on the highest grade completed by the primary caregiver (usually the mother). We do not highlight these findings here because they are generally similar to those for subgroups by number of maternal risk factors (completing less than 12th grade is one of the risk factors counted). They suggest that impacts were smaller for the families with the highest and lowest levels of education (Appendix Tables E.VII.11 through E.VII.13).

• Subgroups defined by the primary caregiver’s living arrangements at enrollment (living with spouse, living with other adults, or living alone with children). We do not highlight these findings here because they are generally similar to those for subgroups by number of maternal risk factors (being a single parent living alone is one of the risk factors counted). They suggest that there were no significant impacts on child development outcomes and impacts on parenting outcomes were smaller for the families in which the primary caregiver was married and lived with a spouse and for families in which the mother lived alone with her children, and the impacts were greater among families in which the primary caregiver lived with other adults and her children (Appendix Tables E.VII.14 through E.VII.16).

B. THE IMPORTANCE OF PARTICULAR SUBGROUP FINDINGS FOR PROGRAMS AND POLICY

The evaluation results for some of the specific subgroups of families described above are especially noteworthy, because they show that the Early Head Start research programs had some important impacts among groups of families that are often the focus of special policies and programs. Below, we highlight these findings and discuss their importance in the context of past research.

1. Working With Teenage Parents and Their Children

Teenage childbearing is an important policy concern because it affects not only a mother’s life but also her child’s. Under pre-welfare reform policies, teenage parents were at especially high risk of long-term welfare dependency. Children of teenage parents are more likely than children of older parents to experience poorer health, less stimulating and supportive home environments, abuse and neglect, difficulties in school, teenage parenthood, and incarceration during young adulthood (Maynard 1996).

Although the Early Head Start programs participating in the research were not designed specifically for teenage mothers, they served teenage mothers and had important favorable impacts on the teenage parents and their children that they served. Despite the challenges they reported in serving teenage parents, the Early Head Start research programs were able to provide substantially more services to teenage parents than they would have obtained on their own in their communities. The programs also produced a favorable pattern of impacts on participation in self-sufficiency-oriented activities among teenage parents.

The pattern of Early Head Start impacts on child development and parenting among teenage parents and their children was weaker than that among older parents and their children, but the impacts on teenage parents and their children are notable in comparison with the impacts of other interventions targeting teenage parents. The Early Head Start programs had a favorable impact on the proportion of children of teenage mothers who scored below 85 on the Bayley MDI and children’s social-emotional behavior. The programs also had significant favorable impacts on parent supportiveness and reported spanking by teenage parents. Finally, the Early Head Start programs increased participation in education activities and toward the end of the follow-up period, the programs reduced welfare receipt among teenage mothers. The program impact on subsequent births among teenage parents was not significant, but it was negative.

These impacts compare favorably with those of other large-scale programs for disadvantaged teenage parents. The Teenage Parent Demonstration programs, which aimed to increase self-sufficiency among teenage parents receiving welfare cash assistance by requiring them to participate in self-sufficiency-oriented activities (with financial sanctions if they did not) and provided support services to enable them to do so (but did not provide intensive services directly to children), significantly increased mothers’ participation in education and employment-related activities and increased their child care use for as long as the requirements were in effect. Based on outcomes measured when children were entering elementary school, the programs did not harm the children of the teenage parents they served, nor did they enhance their development and well-being (Kisker, Rangarajan, and Boller 1998). The voluntary New Chance programs provided comprehensive services to improve self-sufficiency among lowincome teenage parents and improve children’s well-being by helping parents arrange appropriate child care, making referrals for health care, and offering parenting education classes. Many sites offered on-site center-based child care. As voluntary programs, the New Chance programs experienced difficulties recruiting and retaining mothers in program services (the average duration of program participation was approximately 6 months). The programs had no long-term impacts on employment, earnings, income, or welfare receipt and had few impacts on parenting or children’s well-being. The evaluation found small negative impacts on children’s social-emotional development, based on mothers’ reports, but no significant impacts on teachers’ assessments of children’s academic performance or school adjustment (Quint, Bos, and Polit 1997).

The Early Head Start impacts on teenage parents and their children also compare favorably with those of other recent smaller-scale programs. Because the nurse home visitation program designed by David Olds and his colleagues targeted disadvantaged first-time parents who were pregnant, many participants were teenage parents. The evaluation of the program in Elmira, New York found that the program increased stimulation of children’s language skills and provision of educationally-stimulating toys, games and reading materials among poor, unmarried teenage parents, but there were no enduring impacts on their children’s intellectual functioning (Olds, Henderson, and Kitzman 1994). The evaluation of the program in Memphis, where twothirds of mothers were teenagers when they enrolled, found no program effects when children were 2 years old on children’s mental development or reported behavioral problems; however, the program increased the responsiveness and communicativeness of children of mothers with low psychological resources (Olds, et al. 1998).8 Both programs reduced rates of subsequent pregnancies, and in Elmira, the program improved life-course outcomes (increased employment and education achievements, and reduced welfare dependence) for teenage parents (Olds et al. 1998). The Teen Parents as Teachers Demonstration, which operated in four sites in California, provided monthly home visits and group meetings through the child’s second birthday, and for a subset of participants, also provided case management services. The demonstration evaluation showed that the programs increased teenage parents’ acceptance of their child’s behavior during the HOME, improved children’s cognitive development according to the mothers’ reports, and reduced opened cases of abuse and neglect, but had no large or consistent impacts on parenting or observed child development (Wagner and Clayton 1999).

The evaluation of Early Head Start suggests that when programs put a high priority on providing intensive services and focus on child development while working with teenage parents on education, employment, and other issues, they can have significant impacts on the children’s progress at the same time that they improve teenage parents’ progress toward economic selfsufficiency.

2. Engaging Depressed Mothers

Mothers who are depressed are an important, policy-relevant group. Children of mothers who are depressed are at greater risk of experiencing behavioral, health, and academic problems than children of mothers who are not depressed (Anthony 1983; Gelfand and Teti 1990). In the NICHD Early Child Care Study, mothers reporting chronic symptoms of depression were least sensitive when observed playing with their children, and children whose mothers reported feeling depressed performed more poorly on cognitive-linguistic functioning measures and were rated as less cooperative and more problematic at age 3 (NICHD Early Child Care Research Network 1999). Others studies have also documented more negative parenting behaviors and fewer positive parenting behaviors among mothers who were depressed (Koblinsky, Randolph, Roberts, Boyer, and Godsey 2000). Other problems such as poverty and low literacy may exacerbate these risks (Ahluwalia, McGroder, Zaslow, and Hair 2001; Petterson and Albers 2001). In the Early Head Start control group, the outcomes of children at age 3 were often less favorable among the children of mothers who were at risk of depression when they enrolled.

The smaller impacts on service use among mothers at risk of depression at enrollment, reflecting the lower likelihood that program mothers in that group received intensive services, confirms that mothers who were at risk of depression were harder to engage in services than mothers who were not at risk of depression. Although program group families who were at risk of depression at enrollment were more likely than program group families who were not at risk to report receiving mental health services (32 compared with 22 percent reported receiving mental health services), the reported levels of receipt of mental health services by control families at risk and not at risk of depression were similar to their program group counterparts, 353 and impacts on receipt of mental health services were not statistically significant for either group. Many program staff reported that mental health services were lacking in their communities and described the difficulties they experienced in trying to link families to needed mental health services. The pattern of impacts suggests that the Early Head Start programs were unable to increase their families’ access to mental health services beyond what they could have obtained on their own in their communities.

Despite the difficulties they experienced in engaging mothers at risk of depression at enrollment, the programs had notable favorable impacts on children’s social-emotional behavior and parenting among families of depressed parents and their children. They increased parents’ supportiveness during play and reduced detachment and negative regard during play. They also reduced reported spanking and increased the extent to which mothers followed bedtime routines with their children. The programs also improved the social-emotional behavior of children of mothers at risk of depression during play and during the puzzle challenge task. In most cases, the effect sizes ranged from .2 to .4.

Program impacts on the mental health of mothers who were at risk of depression when they enrolled were mixed. Although the programs increased ratings of parent-child dysfunctional interaction by mothers at risk of depression, they also significantly reduced the symptoms of depression reported by mothers in the CES-D Short Form administered when children were 3 years old.

These impacts on parenting and child development suggest that Early Head Start was a protective factor in the lives of children of depressed mothers. The Early Head Start programs helped mothers who were at risk of depression improve their parenting behavior and thereby improve their children’s behavior. These impacts are promising because they may have important implications for the children in the future. Evidence is growing that young children’s emotional adjustment is an important predictor of later school success (Raver 2002).

The Early Head Start findings are promising in light of recent evaluations of welfare-towork programs. Several recent evaluations have found that welfare-to-work programs have increased mothers’ depressive symptoms and reduced their feelings of warmth toward their children, and these impacts may have contributed to the unfavorable impacts on children’s behavior problems that were observed (Ahluwalia, McGroder, Zaslow, and Hair 2001).

Welfare-to-work programs have consistently had no impacts on employment and earnings among the most-depressed enrollees (Michalopoulos and Schwartz 2001). Thus, it is not surprising that the Early Head Start programs also had no impacts on self-sufficiency-oriented outcomes of mothers who were at risk of depression when they enrolled.

The Early Head Start evaluation suggests that efforts to engage mothers who are at risk of depression in intensive services and focusing on child development while working with mothers on their own needs and goals can have significant impacts on parenting and children’s socialemotional behavior at the same time that they appear to improve aspects of the parents’ mental health. The potential for improving mothers’ mental health may be even greater if Early Head Start programs are able to help depressed parents gain better access to mental health services in the community.

3. Working with High-Risk Families

The impact findings suggest that Early Head Start’s potential for making a difference appears to be greatest among families in the middle of the range of demographic risk factors that we measured. Impacts tended to be unfavorable among the small group of families with the highest number of risk factors. It is possible that the services provided by the Early Head Start programs—primarily weekly home visits or regular attendance at centers—were not sufficient to meet the needs of these families, and program expectations for participation may have added to the challenges these parents faced.

The general pattern of impacts by number of maternal risk factors is similar to patterns that have been observed in the past. Other studies examining risk factors and children’s development have also found unfavorable outcomes among children in families with four or more risk factors (Jones, Forehand, Brody, and Armistead 2002; Rutter 1979; and Liaw and Brooks-Gunn 1995). Past evaluations of welfare employment interventions have found the largest impacts among moderately disadvantaged subgroups and smaller and fewer impacts among both less and more disadvantaged sample members, although in a recent analysis of subgroup impacts among 20 welfare-to-work programs, impacts were more similar among less- and more-disadvantaged subgroups (Michalopoulos and Schwartz 2001). More recently, a major life change hypothesis has been suggested as an explanation for unfavorable impacts on high-risk families in previous research (Zaslow et al. 2002; and Zaslow and Eldred 1998). It may offer one possible explanation for the negative impacts among families with the most risk factors in Early Head Start This hypothesis suggests that low-income families who have experienced high levels of instability, change, and risk may be overwhelmed by the changes that a new program introduces into their lives, even though the program is designed to help. As a result, the program requirements may create unintended negative consequences for these families. In addition, Early Head Start families with the most risk factors tended to be in later or incompletely implemented programs, some of which had high initial rates of staff turnover that may have exacerbated change and other difficult circumstances in their families’ lives.

C. CONCLUSIONS AND IMPLICATIONS

The analyses of impacts among subgroups of children and families show that:

• Program impacts on receipt of services were broad-based. The Early Head Start programs substantially increased the receipt of child development and other services in all subgroups of families.

• Children and parents in most subgroups benefited in some way from the programs, but the extent and magnitude of impacts varied. When children were 3 years old, some impacts on child development and parenting were significant in nearly all of the subgroups examined. For some subgroups, such as families with two or fewer demographic risk factors, the significant impacts were limited in size and number, while for other subgroups, such as African American families, significant impacts emerged in multiple areas and were larger in magnitude, with most effect sizes in the 20 to 50 percent range.

• Earlier intervention is better. The subgroup analyses suggest that it is advantageous to enroll families before their child is born and maximize the time available to work with parents and children. The Early Head Start research programs appear to have been more effective in improving child outcomes in families who enrolled before their child was born than in families who enrolled after their child was born. The programs’ impact on cognitive development at age 3 appears to be larger among children who were not yet born at enrollment, and the impacts on children’s social-emotional development are more numerous and larger in this group. However, children who were born at enrollment also benefited from the program, and program impacts on parenting were more similar across these groups.

• Both firstborn and later-born children benefited from participating in Early Head Start. The impacts on some parenting outcomes did not differ significantly between parents of firstborn and later-born children, but the impacts on daily reading to children and discipline strategies were concentrated among parents of later-born children.

• Early Head Start appears to have provided a safety net for children’s development among some groups of families in which parents may have been struggling with their own economic and developmental needs. Like other programs designed to increase self-sufficiency among disadvantaged teenage parents, the Early Head Start programs succeeded in increasing rates of school attendance among teenage mothers. Unlike other large-scale programs, however, the Early Head Start programs also enhanced their children’s development. In the eight sites where data on depressive symptoms were collected at the time of enrollment, the Early Head Start programs had a consistent pattern of favorable impacts on parenting and children’s social-emotional development in families in which the primary caregiver was at risk of depression at enrollment.

• The Early Head Start programs were especially effective in improving child development and parenting outcomes of the African American parents and children who participated and also had a favorable pattern of impacts on the Hispanic parents and children who participated. Although other unmeasured family characteristics may account for the stronger pattern of impacts among African Americans, it appears that the Early Head Start services may have been considerably more effective than other child development and family support services that African American control group families received in their communities. Given the relatively high levels of child care use, including use of center-based care, among African American families, the stronger pattern of impacts on children in this subgroup may in part reflect differences in the quality of infant and toddler child care that program- and control-group children received. The relatively poorer circumstances experienced by African American children and families in the control group (relative to control families in other groups) also may have set the stage for the programs to have a larger impact on this group. The substantial impacts on service receipt may account for the favorable impacts among Hispanic families. The lack of impacts on white children and parents may reflect, in part, the lower likelihood that these families remained enrolled in the program for at least two years.

• Families with multiple risks usually pose difficult challenges for early intervention and family support programs, and this was true for the Early Head Start programs as well. The lack of favorable impacts among families with four or five of the five demographic risk factors we counted suggests that programs may need to reconsider the mix and intensity of services that could help in working with these families. Because families with four or five risk factors were relatively more likely to be in programs that were not fully implemented early, one important focus of efforts to serve the highest-risk families effectively may be full implementation of the Head Start Program Performance Standards.
  

  TABLES

  
  
  TABLE VII.1 IMPACTS ON SERVICE RECEIPT DURING THE FIRST 28 MONTHS, BY PREGNANCY STATUS AT ENROLLMENT (Click on this link to view table)

  TABLE VII.2 IMPACTS ON CHILD AND FAMILY OUTCOMES AT AGE 3, BY PREGNANCY STATUS AT ENROLLMENT (Click on this link to view table)

  TABLE VII.3 IMPACTS ON SELF-SUFFICIENCY AT 28 MONTHS, BY PREGNANCY STATUS AT ENROLLMENT (Click on this link to view table)

  TABLE VII.4 IMPACTS ON SERVICE RECEIPT DURING THE FIRST 28 MONTHS, BY FOCUS CHILD’S BIRTH ORDER (Click on this link to view table)

  TABLE VII.5 IMPACTS ON CHILD AND FAMILY OUTCOMES AT AGE 3, BY FOCUS CHILD’S BIRTH ORDER (Click on this link to view table)

  TABLE VII.6 IMPACTS ON SELF-SUFFICIENCY AT 28 MONTHS, BY FOCUS CHILD’S BIRTH ORDER (Click on this link to view table)

  TABLE VII.7 IMPACTS ON SERVICE RECEIPT DURING THE FIRST 28 MONTHS, BY AGE OF MOTHER AT BIRTH OF FOCUS CHILD (Click on this link to view table)

  TABLE VII.8 IMPACTS ON CHILD AND FAMILY OUTCOMES AT AGE 3, BY AGE OF MOTHER AT BIRTH OF FOCUS CHILD (Click on this link to view table)

  TABLE VII.9 IMPACTS ON SELF-SUFFICIENCY AT 28 MONTHS, BY AGE OF MOTHER AT BIRTH OF FOCUS CHILD (Click on this link to view table)

  TABLE VII.10 IMPACTS ON SERVICE RECEIPT DURING THE FIRST 28 MONTHS, BY RACE/ETHNICITY (Click on this link to view table)

  TABLE VII.11 IMPACTS ON CHILD AND FAMILY OUTCOMES AT AGE 3, BY RACE/ETHNICITY (Click on this link to view table)

  TABLE VII.12 IMPACTS ON SELF-SUFFICIENCY AT 28 MONTHS, BY RACE/ETHNICITY (Click on this link to view table)

  TABLE VII.13 IMPACTS ON SERVICE RECEIPT DURING THE FIRST 28 MONTHS, BY MAIN LANGUAGE SPOKEN AT HOME (Click on this link to view table)

  TABLE VII.14 IMPACTS ON CHILD AND FAMILY OUTCOMES AT AGE 3, BY MAIN LANGUAGE SPOKEN AT HOME (Click on this link to view table)

  TABLE VII.15 IMPACTS ON SELF-SUFFICIENCY AT 28 MONTHS, BY MAIN LANGUAGE SPOKEN AT HOME (Click on this link to view table)

  TABLE VII.16 IMPACTS ON SERVICE RECEIPT DURING THE FIRST 28 MONTHS, BY NUMBER OF MATERNAL RISK FACTORS (Click on this link to view table)

  TABLE VII.17 IMPACTS ON CHILD AND FAMILY OUTCOMES AT AGE 3, BY NUMBER OF MATERNAL RISK FACTORS (Click on this link to view table)

  TABLE VII.18 IMPACTS ON SELF-SUFFICIENCY AT 28 MONTHS, BY NUMBER OF MATERNAL RISK FACTORS (Click on this link to view table)

  TABLE VII.19 IMPACTS ON SERVICE RECEIPT DURING THE FIRST 28 MONTHS, BY MOTHER’S RISK OF DEPRESSION AT ENROLLMENT (Click on this link to view table)

  TABLE VII.20 IMPACTS ON CHILD AND FAMILY OUTCOMES AT AGE 3, BY MOTHER’S RISK OF DEPRESSION AT ENROLLMENT (Click on this link to view table)

  TABLE VII.21 IMPACTS ON SELF-SUFFICIENCY AT 28 MONTHS, BY MOTHERS RISK OF DEPRESSION AT ENROLLMENT (Click on this link to view table)

---

1 We examined the programs’ impacts on 27 subgroups, which were defined based on family characteristics at the time of random assignment. The subgroups were defined based on one characteristic at a time, and these subgroupings naturally overlap. In sensitivity analyses we found that the patterns of differential impacts largely remained after potential confounding characteristics were controlled.(back)

2 Appendix Tables E.IV.1 and E.IV.2 show the configuration of family characteristics across the research sites.(back)

3Appendix Table E.VII.1 describes the overlap in subgroups.(back)

4In discussion with program directors about the process of transitioning families out of Early Head Start when their children were nearing 3 years old, we learned that some families were distressed about having to leave and did not respond to transition planning as anticipated. It is possible that these families were more likely to be those who had been in Early Head Start since before their child was born.(back)

5Children who spoke English as the primary language in the home were assessed using the PPVT; children who spoke Spanish as the primary language in the home were assessed using the Teste de Vocabulario en Images Peabody (TVIP), the Spanish-language version of the PPVT. Among the subgroup of Hispanic children, 90 were assessed using the PPVT and 174 were assessed using the TVIP.(back)

6The requirement of 10 program and 10 control families in the subgroup causes six sites to drop out of the analyses for African American families, nine sites to drop out of the analyses for Hispanic families, and five sites to drop out of the analyses for white families.(back)

7 Discussions with program directors suggest that the programs took steps to make sure that the highest-risk families received services to meet their basic needs and had a “safety net” under them.(back)

8The Early Head Start programs had significant favorable impacts on the social-emotional behavior of children of teenage mothers when the children were 2 years old (ACYF 2001a).(back)

 

 

Table of Contents | Previous | Next